Small molecule drugs are the cornerstone of modern medical practice. However, their use is plagued by the onset of unexpected side effects, often seen only in late-stage clinical trials or after release to the market. As a result, there have bee a number of high profile drug withdrawals and a dearth of new drug development. Characterizing the combinatorial effects of drug treatment is of particular concern. It is very difficult to empirically study these interactions before drugs enter the market because of the small samples of co- prescribed drugs in most late stage clinical drug (Phase III) studies. Some interactions can be predicted based on knowledge of shared pathways of metabolism, but many are idiosyncratic and difficult to predict. Thus, we must create surveillance methods to detect unexpected drug effects and interactions that leverage the power of large-scale clinical databases such as the electronic health records. Mining of electronic health record data for the purpose of identifying adverse drug effects is an increasingly important research challenge. For example, in response to a congressional mandate the Food and Drug Administration (FDA) established the mini-sentinel initiative in 2009 -- a pilot study that links claims and administratve data from over 31 institutions for the purpose of monitoring drug safety surveillance. In addition, public-private partnerships (e.g. the Observational Medical Outcomes Partnership) have sprouted to establish data management and analysis standards for safety surveillance. However, the potential of the EHR for drug surveillance is paralleled by an equal number of challenges. Many of these challenges are in the quality (or rather lack thereof) of data when used for secondary analyses. Data stored in the EHR are often dirty, noisy, and missing. In addition to issues regarding data capture, these data also suffer from bias which confounds analysis and makes data mining results difficult to interpret. These issues become especially acute in the context of combination therapies where the exposed patient cohorts are often small and suffer from unknown (i.e. unstudied) biases. In this proposal we present a drug safety surveillance strategy which integrates state-of-the-art signal detection algorithms with chemical systems biology data for the purpose of identifying unexpected effects of combination therapies. We present an integrative methodology which combines quantitative signal detection and chemical systems biology to mine drug effects from a large clinical database. This will require innovations in observational statistical data mining, network analysis, and integrative chemical systems biology. The result will be a set of tools for discovering drug effects and linking them to molecular interaction networks. These resources will aid federal regulators to better monitor the safety of drugs at the population level, pharmacologists who wish to understand the effects of drugs at the physiological level, and drug development researchers to explore new treatments of human disease.